AI Paper Atlas

Research map / Architectures and efficiency / Quantization

MOSS: Efficient and Accurate FP8 LLM Training with Microscaling and Automatic Scaling

Yu Zhang, Hui-Ling Zhen, Mingxuan Yuan, Bei Yu

ICLR 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of LLMs · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
FP8 training with MOSS instead of BF16Qwen-3-14B/32B · MATH500, GPQA, C-Eval, AIME24, MMLU-Reduxaccuracy: 96.8 -> 96.7 (14B MATH500)No clear effect
use two-level microscaling instead of per-group quantization (ablation)MOSS · activation SNR (LLaMA-2-7B)SNR (dB): 34.5 -> 37.5 early; 32.9 -> 36.0 lateImproved
use two-level microscaling instead of per-tensor quantization (ablation)MOSS · activation SNR (LLaMA-2-7B)SNR (dB): 28.3 -> 37.5 earlyImproved
use MOSS kernel instead of COAT per-group GEMM (ablation)MOSS · FP8 GEMM on H800avg runtime (ms): 3.73 -> 0.77Improved
MOSS GEMM vs DeepGEMMMOSS · FP8 GEMM on H800avg runtime (ms): 0.54 -> 0.77Hurt
use automatic scaling instead of just-in-time scaling (ablation)MOSS · LLaMA-2-7B fine-tuningthroughput (tokens/s): 38,642 -> 41,998 (1.087x)Improved
use automatic scaling instead of just-in-time scaling (ablation)MOSS · LLaMA-2-7B MAmmoTHGSM8K accuracy: 64.9% -> 64.7%No clear effect
increase rescaling interval from 500 to 2000 (ablation)MOSS · LLaMA-2-7B NumGLUEaccuracy: 59.4% -> 58.1%Hurt
use FP8 MOSS instead of COAT for memoryMOSS · peak activation memoryGB: 28.6 -> 23.5Improved
use MOSS FP8 instead of BF16 for pretraining throughputMOSS · OLMo-7Btraining speedup: 1.34x vs BF16Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy